Development Of Machine Learning Models Research Articles

Development of machine learning models to predict papillary carcinoma in thyroid nodules: The role of immunological, radiologic, cytologic and radiomic features

ObjectiveApproximately 30 % of thyroid nodules yield an indeterminate diagnosis through conventional diagnostic strategies. The aim of this study was to develop machine learning (ML) models capable of identifying papillary thyroid carcinomas using preoperative variables. MethodsPatients with thyroid nodules undergoing thyroid surgery were enrolled in a retrospective monocentric study. Six 2-class supervised ML models were developed to predict papillary thyroid carcinoma, by sequentially incorporating clinical-immunological, ultrasonographic, cytological, and radiomic variables. ResultsOut of 186 patients, 92 nodules (49.5 %) were papillary thyroid carcinomas in the histological report. The Area Under the Curve (AUC) ranged from 0.41 to 0.61 using only clinical-immunological variables. All ML models exhibited an increased performance when ultrasound variables were included (AUC: 0.95–0.97). The addition of cytological (AUC: 0.86–0.97) and radiomic (AUC: 0.88–0.97) variables did not further improve ML models’ performance. ConclusionML algorithms demonstrated low accuracy when trained with clinical-immunological data. However, the inclusion of radiological data significantly improved the models' performance, while cytopathological and radiomics data did not further improve the accuracy. Level of evidenceLevel 4.

The Usability and Feasibility of a Dietary Intake Self-Monitoring Application in a Population with Varying Literacy Levels.

Our aim was to study how hemodialysis patients with varying levels of literacy would use a diet and fluid intake monitoring mobile application (DIMA-P) and what would be its impact on their dietary behaviors. We developed a mobile application using user-centered methods and informed by the Integrated Theory of Health Behavior Change (ITHBC). Eight hemodialysis patients were recruited to use the application to record and monitor their diet and fluid intakes for a 6-week study. Overall, the participants had low literacy, numeracy, and technical skills. We collected the data on application usage and administered usability and context-of-use questionnaires to gain insights into the participants' interaction with the application. The participants' portion estimation skills and dietary self-regulation self-efficacy were assessed using various tests. In addition, interdialytic weight gain data were collected to assess the impact of app usage on the participants' health outcomes. The application usage patterns varied, with a general trend towards frequent use (n = 5) correlating with engagement in self-monitoring. The participants gave high comprehensibility, user-friendliness, satisfaction, and usefulness ratings, suggesting that the app was well designed and the target users could easily navigate and interact with the features. While the participants improved in estimating portion sizes, the impact on measuring skills was variable. There was also an improvement in the participants' dietary self-regulation self-efficacy post-study. The interdialytic weight gain trends indicated a slight improvement in fluid and diet management. People with different literacy skills can effectively use icon-based interfaces for portion size estimation and develop personalized usage patterns to self-regulate their fluid and dietary intakes. Moreover, they can experience an enhancement in their dietary self-efficacy skills by using a mobile application aimed at providing nutritional feedback. Furthermore, this research shows that the constructs of the ITHBC are effective in promoting dietary behavior change in a population with varying literacy skills. The target users can benefit from explicitly visualizing the relationship between their health outcomes and the factors influencing those outcomes. These user ambitions could be supported by developing machine learning models. Future research should also focus on enhancing the mechanisms by which technology can further enhance each component of the ITHBC framework.

Computer Vision Identification of Trachomatous Inflammation-Follicular Using Deep Learning.

Trachoma surveys are used to estimate the prevalence of trachomatous inflammation-follicular (TF) to guide mass antibiotic distribution. These surveys currently rely on human graders, introducing a significant resource burden and potential for human error. This study describes the development and evaluation of machine learning models intended to reduce cost and improve reliability of these surveys. Fifty-six thousand seven hundred twenty-five everted eyelid photographs were obtained from 11,358 children of age 0 to 9 years in a single trachoma-endemic region of Ethiopia over a 3-year period. Expert graders reviewed all images from each examination to determine the estimated number of tarsal conjunctival follicles and the degree of trachomatous inflammation-intense. The median estimate of the 3 grader groups was used as the ground truth to train a MobileNetV3 large deep convolutional neural network to detect cases with TF. The classification model predicted a TF prevalence of 32%, which was not significantly different from the human consensus estimate (30%; 95% confidence interval of difference, -2 to +4%). The model had an area under the receiver operating characteristic curve of 0.943, F1 score of 0.923, 88% accuracy, 83% sensitivity, and 91% specificity. The area under the receiver operating characteristic curve increased to 0.995 when interpreting nonborderline cases of TF. Deep convolutional neural network models performed well at classifying TF and detecting the number of follicles evident in conjunctival photographs. Implementation of similar models may enable accurate, efficient, large-scale trachoma screening. Further validation in diverse populations with varying TF prevalence is needed before implementation at scale.

Predicting the Bioaccessibility of Soil Cd, Pb, and As with Advanced Machine Learning for Continental-Scale Soil Environmental Criteria Determination in China.

Investigating the bioaccessibility of harmful inorganic elements in soil is crucial for understanding their behavior in the environment and accurately assessing the environmental risks associated with soil. Traditional batch experimental methods and linear models, however, are time-consuming and often fall short in precisely quantifying bioaccessibility. In this study, using 937 data points gathered from 56 journal articles, we developed machine learning models for three harmful inorganic elements, namely, Cd, Pb, and As. After thorough analysis, the model optimized through a boosting ensemble strategy demonstrated the best performance, with an average R 2 of 0.95 and an RMSE of 0.25. We further employed SHAP values in conjunction with quantitative analysis to identify the key features that influence bioaccessibility. By utilizing the developed integrated models, we carried out predictions for 3002 data points across China, clarifying the bioaccessibility of cadmium (Cd), lead (Pb), and arsenic (As) in the soils of various sites and constructed a comprehensive spatial distribution map of China using the inverse distance weighting (IDW) interpolation method. Based on these findings, we further derived the soil environmental standards for metallurgical sites in China. Our observations from the collected data indicate a reduction in the number of sites exceeding the standard levels for Cd, Pb, and As in mining/smelting sites from 5, 58, and 14 to 1, 24, and 7, respectively. This research offers a precise and scientific approach for cross-regional risk assessment at the continental scale and lays a solid foundation for soil environmental management.

Computed tomography-based body composition indicative of diabetes after hypertriglyceridemic acute pancreatitis

BackgroundPost‑acute pancreatitis prediabetes/diabetes mellitus (PPDM‑A) is one of the common sequelae of acute pancreatitis (AP). The aim of our study was to build a machine learning (ML)-based prediction model for PPDM-A in hypertriglyceridemic acute pancreatitis (HTGP). MethodsWe retrospectively enrolled 165 patients for our study. Demographic and laboratory data and body composition were collected. Multivariate logistic regression was applied to select features for ML. Support vector machine (SVM), linear discriminant analysis (LDA), and logistic regression (LR) were used to develop prediction models for PPDM-A. Results65 patients were diagnosed with PPDM-A, and 100 patients were diagnosed with non-PPDM-A. Of the 84 body composition-related parameters, 15 were significant in discriminating between the PPDM-A and non-PPDM-A groups. Using clinical indicators and body composition parameters to develop ML models, we found that the SVM model presented the best predictive ability, obtaining the best AUC=0.796 in the training cohort, and the LDA and LR model showing an AUC of 0.783 and 0.745, respectively. ConclusionsThe association between body composition and PPDM-A provides insight into the potential pathogenesis of PPDM-A. Our model is feasible for reliably predicting PPDM-A in the early stages of AP and enables early intervention in patients with potential PPDM-A.

Development and Validation of Machine Learning Models to Predict Postoperative Delirium Using Clinical Features and Polysomnography Variables.

Background: Delirium affects up to 50% of patients following high-risk surgeries and is associated with poor long-term prognosis. This study employed machine learning to predict delirium using polysomnography (PSG) and sleep-disorder questionnaire data, and aimed to identify key sleep-related factors for improved interventions and patient outcomes. Methods: We studied 912 adults who underwent surgery under general anesthesia at a tertiary hospital (2013-2024) and had PSG within 5 years of surgery. Delirium was assessed via clinical diagnoses, antipsychotic prescriptions, and psychiatric consultations within 14 days postoperatively. Sleep-related data were collected using PSG and questionnaires. Machine learning predictions were performed to identify postoperative delirium, focusing on model accuracy and feature importance. Results: This study divided the 912 patients into an internal training set (700) and an external test set (212). Univariate analysis identified significant delirium risk factors: midazolam use, prolonged surgery duration, and hypoalbuminemia. Sleep-related variables such as fewer rapid eye movement (REM) episodes and higher daytime sleepiness were also linked to delirium. An extreme gradient-boosting-based classification task achieved an AUC of 0.81 with clinical variables, 0.60 with PSG data alone, and 0.84 with both, demonstrating the added value of PSG data. Analysis of Shapley additive explanations values highlighted important predictors: surgery duration, age, midazolam use, PSG-derived oxygen saturation nadir, periodic limb movement index, and REM episodes, demonstrating the relationship between sleep patterns and the risk of delirium. Conclusions: The artificial intelligence model integrates clinical and sleep variables and reliably identifies postoperative delirium, demonstrating that sleep-related factors contribute to its identification. Predicting patients at high risk of developing postoperative delirium and closely monitoring them could reduce the costs and complications associated with delirium.

Development and validation of machine learning models for predicting cancer-related fatigue in lymphoma survivors

BackgroundNew cases of lymphoma are rising, and the symptom burden, like cancer-related fatigue (CRF), severely impacts the quality of life of lymphoma survivors. However, clinical diagnosis and treatment of CRF are inadequate and require enhancement. ObjectiveThe main objective of this study is to construct machine learning-based CRF prediction models for lymphoma survivors to help healthcare professionals accurately identify the CRF population and better personalize treatment and care for patients. MethodsA cross-sectional study in China recruited lymphoma patients from June 2023 to March 2024, dividing them into two datasets for model construction and external validation. Six machine learning algorithms were used in this study: Logistic Regression (LR), Random Forest, Single Hidden Layer Neural Network, Support Vector Machine, eXtreme Gradient Boosting, and Light Gradient Boosting Machine (LightGBM). Performance metrics like the area under the receiver operating characteristic (AUROC) and calibration curves were compared. The clinical applicability was assessed by decision curve, and Shapley additive explanations was employed to explain variable significance. ResultsCRF incidence was 40.7 % (dataset I) and 44.8 % (dataset II). LightGBM showed strong performance in training and internal validation. LR excelled in external validation with the highest AUROC and best calibration. Pain, total protein, physical function, and sleep disturbance were important predictors of CRF. ConclusionThe study presents a machine learning-based CRF prediction model for lymphoma patients, offering dynamic, data-driven assessments that could enhance the development of automated CRF screening tools for personalized management in clinical practice.

459 A new approach to measuring dry matter intake, from drylot to pasture: Activities of the alliance for regenerative livestock

Abstract The current state of feed intake measurement for cattle in confinement relies on expensive monitoring systems. There is no effective method for measuring feed intake in grazing animals. To address these limitations, our research aims to tap into the potential of passively collected data, such as data from the Internet of Things (IoT), to address practical challenges in animal agriculture. This includes improving livestock management, optimizing resource allocation, and reducing ecological footprints. The benefits of our research extend beyond improved livestock management to include reducing ecological footprints, such as water, carbon, and environmental footprints. Grazing livestock have a crucial role in nutrient cycling in grasslands and advancing more sustainable livestock production systems. This requires innovation of research methods for quantifying intakes of animals housed in a variety of systems including confinement facilities that lack expensive feed intake systems and for grazing animals. Currently, the industry standard for determining feed intake are equations published in the Nutrient Requirements of Beef Cattle (NASEM, 2016). Our recent efforts using machine learning approaches (Blake et al., 2023) dramatically outperforms the equations. Using the same dataset, the NASEM (2016) approach overestimates intake by 34%. We have developed a predictive algorithm utilizing data collected from animals in confinement where individual feed intake (ground truth) is known. We have developed machine learning models such as Repeated Measures Random Forests (RMRF) and Extreme Gradient Boosting (XGB) to better predict feed intake in confined cattle. We have also used these predictive algorithms to determine feed intake of grazing animals and animals in confinement without feed intake systems and validated the predictive value of the approach.

On the Development of an Acoustic Image Dataset for Unexploded Ordnance Classification Using Front-Looking Sonar and Transfer Learning Methods.

This research aimed to develop a dataset of acoustic images recorded by a forward-looking sonar mounted on an underwater vehicle, enabling the classification of unexploded ordnances (UXOs) and objects other than unexploded ordnance (nonUXOs). The dataset was obtained using digital twin simulations performed in the Gazebo environment utilizing plugins developed within the DAVE project. It consists of 69,444 sample images of 512 × 399 resolution organized in two classes annotated as UXO and nonUXO. The obtained dataset was then evaluated by state-of-the-art image classification methods using off-the-shelf models and transfer learning techniques. The research included VGG16, ResNet34, ResNet50, ViT, RegNet, and Swin Transformer. Its goal was to define a base rate for the development of other specialized machine learning models. Neural network experiments comprised two stages-pre-training of only the final layers and pre-training of the entire network. The experiments revealed that to obtain high accuracy, it is required to pre-train the entire network, under which condition, all the models achieved comparable performance, reaching 98% balanced accuracy. Surprisingly, the highest accuracy was obtained by the VGG model.

Development of Machine Learning Models for Predicting the 1-Year Risk of Reoperation After Lower Limb Oncological Resection and Endoprosthetic Reconstruction Based on Data From the PARITY Trial.

Oncological resection and reconstruction involving the lower extremities commonly lead to reoperations that impact patient outcomes and healthcare resources. This study aimed to develop a machine learning (ML) model to predict this reoperation risk. This study was conducted according to TRIPOD + AI. Data from the PARITY trial was used to develop ML models to predict the 1-year reoperation risk following lower extremity oncological resection and reconstruction. Six ML algorithms were tuned and calibrated based on fivefold cross-validation. The best-performing model was identified using classification and calibration metrics. The polynomial support vector machine (SVM) model was chosen as the best-performing model. During internal validation, the SVM exhibited an AUC-ROC of 0.73 and a Brier score of 0.17. Using an optimal threshold that balances all quadrants of the confusion matrix, the SVM exhibited a sensitivity of 0.45 and a specificity of 0.81. Using a high-sensitivity threshold, the SVM exhibited a sensitivity of 0.68 and a specificity of 0.68. Total operative time was the most important feature for reoperation risk prediction. The models may facilitate reoperation risk stratification, allowing for better patient counseling and for physicians to implement measures that reduce surgical risks.

Development of Lung Cancer Risk Prediction Machine Learning Models for Equitable Learning Health System: Retrospective Study.

A significant proportion of young at-risk patients and nonsmokers are excluded by the current guidelines for lung cancer (LC) screening, resulting in low-screening adoption. The vision of the US National Academy of Medicine to transform health systems into learning health systems (LHS) holds promise for bringing necessary structural changes to health care, thereby addressing the exclusivity and adoption issues of LC screening. This study aims to realize the LHS vision by designing an equitable, machine learning (ML)-enabled LHS unit for LC screening. It focuses on developing an inclusive and practical LC risk prediction model, suitable for initializing the ML-enabled LHS (ML-LHS) unit. This model aims to empower primary physicians in a clinical research network, linking central hospitals and rural clinics, to routinely deliver risk-based screening for enhancing LC early detection in broader populations. We created a standardized data set of health factors from 1397 patients with LC and 1448 control patients, all aged 30 years and older, including both smokers and nonsmokers, from a hospital's electronic medical record system. Initially, a data-centric ML approach was used to create inclusive ML models for risk prediction from all available health factors. Subsequently, a quantitative distribution of LC health factors was used in feature engineering to refine the models into a more practical model with fewer variables. The initial inclusive 250-variable XGBoost model for LC risk prediction achieved performance metrics of 0.86 recall, 0.90 precision, and 0.89 accuracy. Post feature refinement, a practical 29-variable XGBoost model was developed, displaying performance metrics of 0.80 recall, 0.82 precision, and 0.82 accuracy. This model met the criteria for initializing the ML-LHS unit for risk-based, inclusive LC screening within clinical research networks. This study designed an innovative ML-LHS unit for a clinical research network, aiming to sustainably provide inclusive LC screening to all at-risk populations. It developed an inclusive and practical XGBoost model from hospital electronic medical record data, capable of initializing such an ML-LHS unit for community and rural clinics. The anticipated deployment of this ML-LHS unit is expected to significantly improve LC-screening rates and early detection among broader populations, including those typically overlooked by existing screening guidelines.

Machine Learning for Accurate Energy Analysis of Double Beta Decay Events

Point-contact germanium detectors are used to detect and analyze particles and their decay processes. This research focuses on the search for a phenomenon known as neutrinoless double beta decay, where two neutrons within a nucleus transform into two protons, emitting electrons in the process. Unlike standard double beta decay, no neutrinos are emitted, suggesting that the neutrino may act as its own antiparticle, effectively canceling itself out. Detecting this process would provide crucial insights into the nature of neutrino mass. During such an event, the detector registers a spike in energy due to the emitted electrons, producing a step-like pulse in the output channels. The energy of the event is determined by the difference in charge between the ‘top’ and ‘bottom’ steps of the pulse. However, accurately determining the energy is complicated by the presence of a resistor in the detector, which causes the top step to decay exponentially back to a baseline in preparation for the next event. This decay makes it challenging to determine the true energy of the pulse. My work this summer focused on developing machine learning models to reconstruct the step-like nature of pulses from their decayed versions recorded by the detector. This involved constructing deep neural networks trained on inputs (decayed pulses) and corresponding targets (the original step pulses) to learn to accurately reconstruct the target from the input. Initially, the networks were trained on simulated detector data. Once proficient in reconstruction, electronic noise was added to the input pulses to simulate real-world conditions, under which the models were trained to remove both electronic noise and decay. The models were trained to handle pulses with varying decay constants to improve their generalization ability. These neural networks can now be applied to real detector data for accurate energy analysis of double beta decay events.

Regression-classification ensemble machine learning model for loading capacity and bucking mode prediction of cold-formed steel built-up I-section columns

This paper presents a regression-classification ensemble machine learning (ML) model for loading capacity and, in particular, buckling mode prediction with respect to cold-formed steel (CFS) I-section columns. The ML model comprises two sub-models, including the regression sub-model for load capacity prediction and the classification sub-model for buckling mode prediction. A total of 541 experimental and numerical data for CFS built-up I-section columns with varies geometric dimensions, material properties and buckling modes were collected from the literature to construct a dataset. To improve the accuracy and the explainability of the ML model on relative small-scale dataset, physical information-enhanced features based on the knowledge of the direct strength method were adopted as the input features of the ML model, including the yield strength, the elastic local buckling load, the elastic distortional buckling load and the elastic global buckling load, instead of directly using the parameters of geometric dimensions and material properties as the input features. The appropriateness of seven classical machine learning algorithms was compared, namely Decision Tree (DT), Random Forest (RF), Support Vector Machine (SVM), K-Nearest Neighbour (KNN), Adaptive Boosting (ADA), Extreme Gradient Boosting (XG), and Categorical Boosting (CB). The statistical analysis showed that the XG algorithm had the highest level of accuracy for both loading capacity and buckling mode prediction. In addition, the explainability analysis indicated that the developed ML model successfully learned the mechanical characteristic of CFS built-up I-section columns. Finally, the developed ML model was further compared with the existing codified method in the literature. The comparison results indicated that the developed ML model can significantly improve the level of accuracy for loading capacity and buckling mode prediction, which provides a promising alternative choice for the design of CFS built-up I-section columns

Artificial Neural Network and Adaptive Neuro Fuzzy Inference System Hybridized Models in the Sustainable Integration of Language and Mathematics Skills: The Case of Singapore and Hong Kong

The four basic language skills, listening, speaking, reading, and writing play, a crucial role in the development of an individual’s skills in other disciplines. The current study aims to underpin the relationship between language skills and mathematics skills by focusing on the language and mathematics curricula of two consistently high-achieving countries, Hong Kong and Singapore, in the Program for International Student Assessment (PISA) rankings. In the current study, the convergent parallel mixed method was utilized that qualitative and quantitative data were composed together. Primarily, the outcomes of four language skills were determined in the native language teaching curricula of the two countries. The topics and themes related to four basic language skills were determined from the two mathematics curricula. The curricula were examined by document analysis from qualitative research methods. The analysis was conducted by examining the native language teaching and the mathematics curricula of both countries by the content analysis method. Later, the findings of the document analysis were used to develop machine learning models to find a possible positive relationship between language skills and the PISA scores. Although a number of previous studies have found a reasonable relationship between language skills and mathematics skills, the current study results were contradictory to the ones performed previously in the literature, and considering the curricula no positive relationship between the language and mathematics skills was found. The findings of the current study were further supported by the artificial neural network (ANN) and Adaptive Neuro Fuzzy Inference System (ANFIS) model performance metrics. Compared to an acceptable level of 0.80, significantly low R2 values of 0.35 and 0.39 for the ANN and ANFIS models, respectively, indicated very little relationship between the language and mathematics skills.

A Machine Learning Model for Training Your AI

Artificial Intelligence is playing an increasing role in solving some of the world’s biggest problems. Machine Learning Models, within the context of reinforcement learning, define and structure a problem in a format that can be used to learn about an environment in order to find an optimal solution. This includes the states, actions, rewards, and other elements in a learning environment. This also includes the logic and policies that guide learning agents to an optimal or nearly optimal solution to the problem. This paper outlines a process for developing machine learning models. The process is extensible and can be applied to solve various problems. This includes a process for implementing data models using multi-dimensional arrays for efficient data processing. We include an evaluation of learning policies, assessing their performance relative to manual and automated approaches.

Fair swarm learning: Improving incentives for collaboration by a fair reward mechanism

Swarm learning is an emerging technique for collaborative machine learning in which several participants train machine learning models without sharing private data. In a standard swarm network, all the nodes in the network receive identical final models regardless of their individual contributions. This mechanism may be deemed unfair from an economic perspective, discouraging organizations with more resources from participating in any collaboration. Here, we present a framework for swarm learning in which nodes receive personalized models based on their contributions. The results of this study demonstrate the efficacy of this approach by showing that all participants experience performance enhancements compared to their local models. However, participants with higher contributions receive better models than those with lower contributions. This fair mechanism results in the highest possible accuracy for the most contributive participant, comparable to the standard swarm learning model. Such incentive structure can motivate resource-rich organizations to engage in collaboration, leading to the development of machine learning models that incorporate data from more resources, which is ultimately beneficial for every party.

Development and validation of machine learning models for predicting HER2-zero and HER2-low breast cancers.

To develop and validate machine learning models for human epidermal growth factor receptor 2 (HER2)-zero and HER2-low using MRI features pre-neoadjuvant therapy (NAT). Five hundred and sixteen breast cancer patients post-NAT surgery were randomly divided into training (n = 362) and internal validation sets (n = 154) for model building and evaluation. MRI features (tumour diameter, enhancement type, background parenchymal enhancement, enhancement pattern, percentage of enhancement, signal enhancement ratio, breast oedema, and apparent diffusion coefficient) were reviewed. Logistic regression (LR), support vector machine (SVM), k-nearest neighbour (KNN), and extreme gradient boosting (XGBoost) models utilized MRI characteristics for HER2 status assessment in training and validation datasets. The best-performing model generated a HER2 score, which was subsequently correlated with pathological complete response (pCR) and disease-free survival (DFS). The XGBoost model outperformed LR, SVM, and KNN, achieving an area under the receiver operating characteristic curve (AUC) of 0.783 (95% CI, 0.733-0.833) and 0.787 (95% CI, 0.709-0.865) in the validation dataset. Its HER2 score for predicting pCR had an AUC of 0.708 in the training datasets and 0.695 in the validation dataset. Additionally, the low HER2 score was significantly associated with shorter DFS in the validation dataset (hazard ratio: 2.748, 95% CI, 1.016-7.432, P = .037). The XGBoost model could help distinguish HER2-zero and HER2-low breast cancers and has the potential to predict pCR and prognosis in breast cancer patients undergoing NAT. HER2-low-expressing breast cancer can benefit from the HER2-targeted therapy. Prediction of HER2-low expression is crucial for appropriate management. MRI features offer a solution to this clinical issue.

Intelligent decision-making for TBM tunnelling control parameters using multi-objective optimization

In tunnel construction, tunnel boring machine (TBM) tunnelling typically relies on manual experience with sub-optimal control parameters, which can easily lead to inefficiency and high costs. This study proposed an intelligent decision-making method for TBM tunnelling control parameters based on multi-objective optimization (MOO). First, the effective TBM operation dataset is obtained through data preprocessing of the Songhua River (YS) tunnel project in China. Next, the proposed method begins with developing machine learning models for predicting TBM tunnelling performance parameters (i.e. total thrust and cutterhead torque), rock mass classification, and hazard risks (i.e. tunnel collapse and shield jamming). Then, considering three optimal objectives, (i.e., penetration rate, rock-breaking energy consumption, and cutterhead hob wear), the MOO framework and corresponding mathematical expression are established. The Pareto optimal front is solved using DE-NSGA-II algorithm. Finally, the optimal control parameters (i.e., advance rate and cutterhead rotation speed) are obtained by the satisfactory solution determination criterion, which can balance construction safety and efficiency with satisfaction. Furthermore, the proposed method is validated through 50 cases of TBM tunnelling, showing promising potential of application.

OS02-06 Leveraging in silico binding affinity, in vitro bioactivity, and chemical structure to develop machine learning models for predicting in vivo toxicity

OS02-06 Leveraging in silico binding affinity, in vitro bioactivity, and chemical structure to develop machine learning models for predicting in vivo toxicity

Machine learning in paediatric haematological malignancies: a systematic review of prognosis, toxicity and treatment response models.

Machine Learning (ML) has demonstrated potential in enhancing care in adult oncology. However, its application in paediatric haematological malignancies is still emerging, necessitating a comprehensive review of its capabilities and limitations in this area. A literature search was conducted through Ovid. Studies included focused on ML models in paediatric patients with haematological malignancies. Studies were categorised into thematic groups for analysis. Twenty studies, primarily on leukaemia, were included in this review. Studies were organised into thematic categories such as prognoses, treatment responses and toxicity predictions. Prognostic studies showed AUC scores between 0.685 and 0.929, indicating moderate-high predictive accuracy. Treatment response studies demonstrated AUC scores between 0.840 and 0.875, reflecting moderate accuracy. Toxicity prediction studies reported high accuracy with AUC scores from 0.870 to 0.927. Only five studies (25%) performed external validation. Significant heterogeneity was noted in ML tasks, reporting formats, and effect measures across studies, highlighting a lack of standardised reporting and challenges in data comparability. The clinical applicability of these ML models remains limited by the lack of external validation and methodological heterogeneity. Addressing these challenges through standardised reporting and rigorous external validation is needed to translate ML from a promising research tool into a reliable clinical practice component. Key message: Machine Learning (ML) significantly enhances predictive models in paediatric haematological cancers, offering new avenues for personalised treatment strategies. Future research should focus on developing ML models that can integrate with real-time clinical workflows. Addition to literature: Provides a comprehensive overview of current ML applications and trends. It identifies limitations to its applicability, including the limited diversity in datasets, which may affect the generalisability of ML models across different populations. Encourages standardisation and external validation in ML studies, aiming to improve patient outcomes through precision medicine in paediatric haematological oncology.